Protected Audience API developer guide

As you read through the Privacy Sandbox on Android documentation, use the Developer Preview or Beta button to select the program version that you're working with, as instructions may vary.


The Protected Audience API on Android (formerly known as FLEDGE) includes the Custom Audience API and the Ad Selection API. Ad tech platforms and advertisers can use these APIs to serve customized ads based on previous app engagement that limits the sharing of identifiers across apps and limits sharing a user's app interaction information with third-parties.

The Custom Audience API is centered around the "custom audience" abstraction, which represents a group of users with common intentions. An advertiser can register a user with a custom audience and associate relevant ads with it. This information is stored locally and can be used to inform advertiser bids, ad filtering, and ad rendering.

The Ad Selection API provides a framework that allows multiple developers to run an auction locally for a custom audience. To achieve this, the system considers relevant ads associated with the custom audience and performs additional processing on ads that an ad tech platform returns to the device.

Ad tech platforms can integrate these APIs to implement remarketing that preserves user privacy. Support for additional use cases, including app install ads, are planned for future releases. Learn more about the Protected Audience API on Android in the design proposal.

This guide describes how to work with the Protected Audience API on Android to do the following:

  1. Manage custom audiences
  2. Set up and run ad selection on a device
  3. Report ad impressions

Before you begin

Before you get started, complete the following:

  1. Set up your development environment for the Privacy Sandbox on Android.
  2. Either install a system image onto a supported device or set up an emulator that includes support for the Privacy Sandbox on Android.
  3. In a terminal, enable access to the Protected Audience API (disabled by default) with the following adb command.

      adb shell device_config put adservices ppapi_app_allow_list \"*\"
    
  4. In a terminal, enable beacon reporting with the following adb commands.

     adb shell device_config put adservices fledge_beacon_reporting_metrics_enabled true
     adb shell device_config put adservices fledge_register_ad_beacon_enabled true
    
  5. Include an ACCESS_ADSERVICES_CUSTOM_AUDIENCE permission in your app manifest:

      <uses-permission android:name="android.permission.ACCESS_ADSERVICES_CUSTOM_AUDIENCE" />
    
  6. Reference an ad services configuration in the <application> element of your manifest:

      <property android:name="android.adservices.AD_SERVICES_CONFIG"
                android:resource="@xml/ad_services_config" />
    
  7. Specify the ad services XML resource referenced in your manifest, such as res/xml/ad_services_config.xml. Learn more about ad services permissions and SDK access control.

      <ad-services-config>
        <custom-audiences allowAllToAccess="true" />
      </ad-services-config>
    
  8. By default, the Ad Selection API enforces limits on the maximum amount of memory that an auction or impression reporting script can allocate. The memory limitation feature requires WebView version 105.0.5195.58 or higher. The platform enforces a version check and calls to the selectAds and reportImpression APIs fail if this isn't satisfied. There are two options to set this up:

    • Option 1: Run the following adb command to deactivate this check:

      adb device_config put fledge_js_isolate_enforce_max_heap_size false
      
    • Option 2: Install WebView Beta from the Google Play store. This must be equal to or higher than the version stated earlier.

Join a custom audience

A custom audience represents a group of users with common intentions or interests as decided by an advertiser app. An app or SDK may use a custom audience to indicate a particular audience, such as someone who has left items in a shopping cart. To create or join a custom audience asynchronously, do the following:

  1. Initialize the CustomAudienceManager object.
  2. Create a CustomAudience object by specifying key parameters such as the buyer's package and a relevant name. Then, initialize the JoinCustomAudienceRequest object with the CustomAudience object.
  3. Call the asynchronous joinCustomAudience() with the JoinCustomAudienceRequest object and relevant Executor and OutcomeReceiver objects.

Kotlin

val customAudienceManager: CustomAudienceManager =
    context.getSystemService(CustomAudienceManager::class.java)

// Initialize a custom audience.
val audience = CustomAudience.Builder()
    .setBuyer(buyer)
    .setName(name)
    ...
    .build()

// Initialize a custom audience request.
val joinCustomAudienceRequest: JoinCustomAudienceRequest =
    JoinCustomAudienceRequest.Builder().setCustomAudience(audience).build()

// Request to join a custom audience.
customAudienceManager.joinCustomAudience(joinCustomAudienceRequest,
    executor,
    outcomeReceiver)

Java

CustomAudienceManager customAudienceManager =
    context.getSystemService(CustomAudienceManager.class);

// Initialize a custom audience.
CustomAudience audience = new CustomAudience.Builder()
    .setBuyer(buyer)
    .setName(name)
    ...
    .build();

// Initialize a custom audience request.
JoinCustomAudienceRequest joinCustomAudienceRequest =
    new JoinCustomAudienceRequest.Builder().setCustomAudience(audience).build();

// Request to join a custom audience.
customAudienceManager.joinCustomAudience(joinCustomAudienceRequest,
    executor,
    outcomeReceiver);

The combination of the following parameters uniquely identifies each CustomAudience object on a device:

  • owner: Package name of the owner app. This is implicitly set to the package name of the caller app.
  • buyer: Identifier for the buyer ad network which manages ads for this custom audience.
  • name: An arbitrary name or identifier for the custom audience.

Calling joinCustomAudience() repeatedly with a different instance of CustomAudience updates any existing CustomAudience with matching owner, buyer, and name parameters. To help preserve privacy, the result of the API does not distinguish between "creation" and "update."

Additionally, the CustomAudience must be created with these required parameters:

Optional parameters for a CustomAudience object may include:

  • Activation time: A custom audience can only participate in ad selection and daily updates after its activation time. This can be useful to engage lapsed users of an app, for example.
  • Expiration time: A future time after which the custom audience is removed from the device.
  • User bidding signals: A JSON string containing user signals, such as the user's preferred locale, that a buyer's bidding logic JavaScript consumes to generate bids during the ad selection process. This format helps ad tech platforms reuse code across platforms and eases the consumption in JavaScript functions.
  • Trusted bidding data: An HTTPS URL and a list of strings used during the ad selection process that fetch bidding signals from a trusted Key/Value service.
  • Ads: A list of AdData objects corresponding to the ads that participate in ad selection. Each AdData object consists of:
    • Render URL: An HTTPS URL that is queried to render the final ad.
    • Metadata: A JSON object serialized as a string containing information to be consumed by buyer bidding logic during the ad selection process.
    • Ad Filters: A class that contains all necessary information for app install ad filtering and frequency capping during ad selection.

Here's an example of a CustomAudience object instantiation:

Kotlin

// Minimal initialization of a CustomAudience object
val customAudience: CustomAudience = CustomAudience.Builder()
    .setBuyer(AdTechIdentifier.fromString("my.buyer.domain.name"))
    .setName("example-custom-audience-name")
    .setDailyUpdateUrl(Uri.parse("https://DAILY_UPDATE_URL"))
    .setBiddingLogicUrl(Uri.parse("https://BIDDING_LOGIC_URL"))
    .build()

Java

// Minimal initialization of a CustomAudience object
CustomAudience customAudience = CustomAudience.Builder()
    .setBuyer(AdTechIdentifier.fromString("my.buyer.domain.name"))
    .setName("example-custom-audience-name")
    .setDailyUpdateUrl(Uri.parse("https://DAILY_UPDATE_URL"))
    .setBiddingLogicUrl(Uri.parse("https://BIDDING_LOGIC_URL"))
    .build();

Handle joinCustomAudience() outcomes

The asynchronous joinCustomAudience() method uses the OutcomeReceiver object to signal the result of the API call.

  • The onResult() callback signifies the custom audience is successfully created or updated.
  • The onError() callback signifies two possible conditions.

Here's an example of handling the outcome of joinCustomAudience():

Kotlin

var callback: OutcomeReceiver<Void, AdServicesException> =
    object : OutcomeReceiver<Void, AdServicesException> {
    override fun onResult(result: Void) {
        Log.i("CustomAudience", "Completed joinCustomAudience")
    }

    override fun onError(error: AdServicesException) {
        // Handle error
        Log.e("CustomAudience", "Error executing joinCustomAudience", error)
    }
};

Java

OutcomeReceiver callback = new OutcomeReceiver<Void, AdServicesException>() {
    @Override
    public void onResult(@NonNull Void result) {
        Log.i("CustomAudience", "Completed joinCustomAudience");
    }

    @Override
    public void onError(@NonNull AdServicesException error) {
        // Handle error
        Log.e("CustomAudience", "Error executing joinCustomAudience", error);
    }
};

Leave a custom audience

If the user no longer satisfies the business criteria for a given custom audience, an app or SDK can call leaveCustomAudience() to remove the custom audience from the device. To remove a CustomAudience based on its unique parameters, do the following:

  1. Initialize the CustomAudienceManager object.
  2. Initialize the LeaveCustomAudienceRequest with the custom audience's buyer and name. To learn more about these input fields, read "Join a custom audience."
  3. Call the asynchronous leaveCustomAudience() method with the LeaveCustomAudienceRequest object and relevant Executor and OutcomeReceiver objects.

Kotlin

val customAudienceManager: CustomAudienceManager =
    context.getSystemService(CustomAudienceManager::class.java)

// Initialize a LeaveCustomAudienceRequest
val leaveCustomAudienceRequest: LeaveCustomAudienceRequest =
    LeaveCustomAudienceRequest.Builder()
        .setBuyer(buyer)
        .setName(name)
        .build()

// Request to leave a custom audience
customAudienceManager.leaveCustomAudience(
    leaveCustomAudienceRequest,
    executor,
    outcomeReceiver)

Java

CustomAudienceManager customAudienceManager =
    context.getSystemService(CustomAudienceManager.class);

// Initialize a LeaveCustomAudienceRequest
LeaveCustomAudienceRequest leaveCustomAudienceRequest =
    new LeaveCustomAudienceRequest.Builder()
        .setBuyer(buyer)
        .setName(name)
        .build();

// Request to leave a custom audience
customAudienceManager.leaveCustomAudience(
    leaveCustomAudienceRequest,
    executor,
    outcomeReceiver);

Similar to calling joinCustomAudience(), the OutcomeReceiver signals the end of an API call. To help protect privacy, an error outcome doesn't distinguish between internal errors and invalid arguments. The onResult() callback is called when the API call has completed, whether or not a matching custom audience is removed successfully.

Run ad selection

To use the Protected Audience API to select ads, call the selectAds() method:

  1. Initialize an AdSelectionManager object.
  2. Build an AdSelectionConfig object.
  3. Call the asynchronous selectAds() method with the AdSelectionConfig object and relevant Executor and OutcomeReceiver objects.

Kotlin

val adSelectionManager: AdSelectionManager =
  context.getSystemService(AdSelectionManager::class.java)

// Initialize AdSelectionConfig
val adSelectionConfig: AdSelectionConfig =
  AdSelectionConfig.Builder().setSeller(seller)
    .setDecisionLogicUrl(decisionLogicUrl)
    .setCustomAudienceBuyers(customAudienceBuyers)
    .setAdSelectionSignals(adSelectionSignals)
    .setSellerSignals(sellerSignals)
    .setPerBuyerSignals(perBuyerSignals)
    .setBuyerContextualAds(
      Collections.singletonMap(
        contextualAds.getBuyer(), contextualAds
      )
    ).build()

// Run ad selection with AdSelectionConfig
adSelectionManager.selectAds(
  adSelectionConfig, executor, outcomeReceiver
)

Java

AdSelectionManager adSelectionManager =
    context.getSystemService(AdSelectionManager.class);

// Initialize AdSelectionConfig
AdSelectionConfig adSelectionConfig =
  new AdSelectionConfig.Builder()
    .setSeller(seller)
    .setDecisionLogicUrl(decisionLogicUrl)
    .setCustomAudienceBuyers(customAudienceBuyers)
    .setAdSelectionSignals(adSelectionSignals)
    .setSellerSignals(sellerSignals)
    .setPerBuyerSignals(perBuyerSignals)
    .setBuyerContextualAds(
      Collections.singletonMap(contextualAds.getBuyer(), contextualAds)
    )
    .build();

// Run ad selection with AdSelectionConfig
adSelectionManager.selectAds(adSelectionConfig, executor, outcomeReceiver);

The selectAds() method requires an AdSelectionConfig input, where you must specify the following required parameters:

  • Seller: Identifier for the seller ad network initiating the ad selection.
  • Decision logic URL: An HTTPS URL queried to obtain the seller ad network's JavaScript logic.
    • HTTPS URL: queried to obtain the seller ad network's JavaScript logic. See the required function signatures.
    • Prebuilt URI: that follows FLEDGE's ad selection format. IllegalArgumentException is thrown, if an unsupported or malformed prebuilt uri is passed.
  • Custom audience buyers: A full list of identifiers for buyer ad networks that are allowed by the seller to participate in the ad selection process. These buyer identifiers correspond to CustomAudience.getBuyer() of participating custom audiences.

The following parameters can be optionally specified for more customized ad selection:

  • Ad selection signals: A JSON object, serialized as a string, containing signals to be consumed by buyer bidding logic JavaScript fetched from CustomAudience.getBiddingLogicUrl().
  • Seller signals: A JSON object, serialized as a string, containing signals consumed by the seller's fetched JavaScript decision logic from AdSelectionConfig.getDecisionLogicUrl().
  • Per buyer signals: A map of JSON objects, serialized as strings, containing signals to be consumed by specific buyers' bidding logic JavaScript fetched from CustomAudience.getBiddingLogicUrl(), which are identified by the buyer fields of participating custom audiences.
  • Contextual ads: A collection of ad candidates that are collected directly from buyers during an auction that happens outside of a Protected Audience auction.

Once an ad is selected, the results, bids, and signals are persisted internally for reporting. The OutcomeReceiver.onResult() callback returns an AdSelectionOutcome that contains:

  • A render URL for the winning ad, obtained from AdData.getRenderUrl().
  • An ad selection ID unique to the device user. This ID is used for reporting the ad impression.

If the ad selection can't be completed successfully due to reasons such as invalid arguments, timeouts, or excessive resource consumption, the OutcomeReceiver.onError() callback provides an AdServicesException with the following behaviors:

  • If the ad selection is initiated with invalid arguments, the AdServicesException indicates an IllegalArgumentException as the cause.
  • All other errors receive an AdServicesException with an IllegalStateException as the cause.

Contextual ads

Protected Audience can incorporate contextual ads into a Protected Auction. Contextual ads need to be selected on the ad tech server and returned to the device outside of Protected Audience APIs. Contextual ads can then be included in the auction using the AdSelectionConfig at which point they function the same as on device ads, including eligibility for negative ad filtering. Once the Protected Audience auction has completed, you need to invoke reportImpression(). This calls reportWin() in the winning contextual ad, in the same pattern as impression reporting, to receive the winning ad on a device. Each contextual ad needs a buyer, a bid, a link to reporting logic, a render URL, and ad metadata.

To deploy contextual ads in app, the target app needs to create a ContextualAds object:

Kotlin

val contextualAds: ContextualAds =
  Builder().setBuyer(AdTechIdentifier.fromString(mBiddingLogicUri.getHost()))
    //Pass in your valid app install ads
    .setDecisionLogicUri(mContextualLogicUri)
    .setAdsWithBid(appInstallAd)
    .build()

Java

ContextualAds contextualAds = new ContextualAds.Builder()
  .setBuyer(AdTechIdentifier.fromString(mBiddingLogicUri.getHost()))
  .setDecisionLogicUri(mContextualLogicUri)
  //Pass in your valid app install ads
  .setAdsWithBid(appInstallAd)
  .build();

The resulting ContextualAds object can then be passed along when creating your AdSelectionConfig:

Kotlin

// Create a new ad
val noFilterAd: AdData = Builder()
  .setMetadata(JSONObject().toString())
  .setRenderUri(Uri.parse(baseUri + NO_FILTER_RENDER_SUFFIX))
  .build()
val noFilterAdWithBid = AdWithBid(noFilterAd, NO_FILTER_BID)
contextualAds.getAdsWithBid().add(noFilterAdWithBid)

Java

// Create a new ad
AdData noFilterAd = new AdData.Builder()
  .setMetadata(new JSONObject().toString())
  .setRenderUri(Uri.parse(baseUri + NO_FILTER_RENDER_SUFFIX))
  .build();
AdWithBid noFilterAdWithBid = new AdWithBid(noFilterAd, NO_FILTER_BID);
contextualAds.getAdsWithBid().add(noFilterAdWithBid);

App install ad filtering

App install ads filtering helps you to filter installation ads for apps that are already installed on a device.

The first step in this process is to define which advertisers have the ability to filter on the installed package. This needs to happen in the app you want to target with an ad.

Kotlin

//Create a request for setting the app install advertisers
val adtech = AdTechIdentifier.fromString("your.enrolled.uri")
val adtechSet = setOf(adtech)
val request = SetAppInstallAdvertisersRequest(adtechSet)

//Set the app install advertisers in the ad selection manager
mAdSelectionManager.setAppInstallAdvertisers(
  request,
  mExecutor,
  object : OutcomeReceiver<Any?, Exception?>() {
    fun onResult(@NonNull ignoredResult: Any?) {
      Log.v("[your tag]", "Updated app install advertisers")
    }

    fun onError(@NonNull error: Exception?) {
      Log.e("[your tag]", "Failed to update app install advertisers", error)
    }
  })

Java

//Create a request for setting the app install advertisers
AdTechIdentifier adtech = AdTechIdentifier.fromString("your.enrolled.uri");
Set<AdTechIdentifier> adtechSet = Collections.singleton(adtech);
SetAppInstallAdvertisersRequest request = new SetAppInstallAdvertisersRequest(adtechSet);

//Set the app install advertisers in the ad selection manager
mAdSelectionManager.setAppInstallAdvertisers(
  request,
  mExecutor,
  new OutcomeReceiver<Object, Exception>() {
    @Override
    public void onResult(@NonNull Object ignoredResult) {
      Log.v("[your tag]", "Updated app install advertisers");
    }

    @Override
    public void onError(@NonNull Exception error) {
      Log.e("[your tag]", "Failed to update app install advertisers", error);
    }
  });

When the preceding code executes, the advertisers passed in are then able to filter out the installed apps that you specify during their bid generation. If you need to remove an advertiser from having access to this app's install status, run this code again with the advertiser's information removed.

The next step is to set up ad filtering inside the publisher app. The party that serves the ad inside of the publisher app (most likely to be a supply-side SDK) must initialize their AdFilters object with information about which ads related to apps they'd like to filter out:

Kotlin

// Instantiate AdFilters object with package names.
val filters: AdFilters = Builder().setAppInstallFilters(
    Builder().setPackageNames(setOf("example.target.app")).build()
  ).build()

Java

// Instantiate AdFilters object with package names.
AdFilters filters = new AdFilters.Builder()
.setAppInstallFilters(
  new AppInstallFilters.Builder()
  .setPackageNames(Collections.singleton("example.target.app"))
  .build())
.build();

Demand-side publishers may also set an AdFilter for ads that exist inside of their custom audiences.

AdFilters can also be passed in at the point of instantiating a new AdData object:

Kotlin

// Instantiate an AdData object with the AdFilters created in the
// previous example.
val appInstallAd: AdData =
  Builder().setMetadata("{ ... }") // Valid JSON string
    .setRenderUri(Uri.parse("www.example-dsp1.com/.../campaign123.html"))
    .setAdFilters(filters).build()

Java

// Instantiate an AdData object with the AdFilters created in the
// previous example.
AdData appInstallAd = new AdData.Builder()
.setMetadata("{ ... }") // Valid JSON string
.setRenderUri(Uri.parse("www.example-dsp1.com/.../campaign123.html"))
    .setAdFilters(filters)
    .build();

Frequency cap filtering

Frequency cap filtering enables ad techs to limit the number of times an ad is shown. Frequency cap filtering reduces ad overexposure and optimizes alternate ad selection for a given ad campaign.

There are two main components of a frequency cap filter: the ad event type, and the ad counter key. The available ad event types that can be used are:

  • Win: A win event indicates that the ad has won an auction. Win events are automatically updated by the Protected Audience API and cannot be called directly by the developer. Win data is only visible to ads within a given custom audience.
  • Impression: Separate from reportImpression, an on-device caller (SSP or MMP) uses updateAdCounterHistogram() to invoke impression events at the point in code they choose. Impression events are visible to all ads belonging to a given DSP, and are not limited to ads in the same custom audience.
  • View: Event is invoked by the on-device caller (SSP or MMP) at a point in code they choose using a call to updateAdCounterHistogram(). View events are visible to all ads belonging to a given DSP and not limited to ads in the same Custom Audience.
  • Click: Event is invoked by the on-device caller (SSP or MMP) at a point in code they choose using a call to updateAdCounterHistogram(). Click events are visible to all ads belonging to a given DSP and not limited to ads in the same Custom Audience.

In the publisher app, an SSP or MMP that has a presence on the device invokes ad events. When updateAdCounterHistogram() is called, the counter of a frequency cap filter is incremented so that future auctions will have up to date information about a user's exposure to a given ad. The ad event types are not forcibly tied to the corresponding user action and are guidelines given to help callers to structure their event system. To increment ad counters at the time of an event, the on-device actor provides the winning ad auction's ad selection ID.

Ad counter keys are arbitrary 32-bit signed integers assigned by a buyer ad tech, and they correspond to a given set of ads as defined by the DSP. Since ad counter keys are limited only to ads that belong to a given DSP, these keys can be selected without overlapping with histograms from another ad tech. Ad counter keys are used to increment DSP-specific identifiers across a DSP's ads or within a given custom audience to filter out ads from future auctions.

Counter keys can be leveraged to prioritize ads that are more likely to be interesting to a given user based on their interactions with other ads from a given buyer ad tech. For example, an ad that has received a high level of engagement from winning ad auctions, views, and clicks, represents an inferred data point. To further illustrate this point: an ad for left handed golf clubs might indicate that the user wouldn't be interested in right handed ones. A frequency cap filter set for a counter key assigned to left-handed ads could filter out ads for right handed clubs.

To use frequency capping in your auction, you must first create KeyedFrequencyCap objects as shown below:

Kotlin

// Value used when incrementing frequency counter
val adCounterKey = 123

// Frequency cap exceeded after 2 counts
val keyedFrequencyCapForImpression: KeyedFrequencyCap = Builder(
  adCounterKey, 2, Duration.ofSeconds(10)
).build()

// Frequency cap exceeded after 1 counts
val keyedFrequencyCapForImpression: KeyedFrequencyCap = Builder(
  adCounterKey, 1, Duration.ofSeconds(10)
).build()

Java

// Value used when incrementing frequency counter
int adCounterKey = 123;

// Frequency cap exceeded after 2 counts
KeyedFrequencyCap keyedFrequencyCapForImpression =
  new KeyedFrequencyCap.Builder(
    adCounterKey, 2, Duration.ofSeconds(10)
  ).build();

// Frequency Cap exceeded after 1 counts
KeyedFrequencyCap keyedFrequencyCapForClick =
  new KeyedFrequencyCap.Builder(
    adCounterKey, 1, Duration.ofSeconds(10)
  ).build();

Once the KeyedFrequencyCap objects are created, you can pass them into an AdFilters object.

Kotlin

val filters: AdFilters = Builder()
  .setFrequencyCapFilters(
    Builder()
      .setKeyedFrequencyCapsForImpressionEvents(
        ImmutableObject.of(keyedFrequencyCapForImpression)
      )
      .setKeyedFrequencyCapsForClickEvents(
        ImmutableObject.of(keyedFrequencyCapForClick)
      )
  ).build()

Java

AdFilters filters = new AdFilters.Builder()
    .setFrequencyCapFilters(new FrequencyCapFilters.Builder()
        .setKeyedFrequencyCapsForImpressionEvents(
            ImmutableObject.of(keyedFrequencyCapForImpression)
        )
        .setKeyedFrequencyCapsForClickEvents(
            ImmutableObject.of(keyedFrequencyCapForClick)
        )
    ).build();

When the AdFilters object is populated with frequency cap filters, it can be passed along when the custom audience is created:

Kotlin

// Initialize a custom audience.
val audience: CustomAudience = Builder()
  .setBuyer(buyer)
  .setName(name)
  .setAds(
    listOf(
      Builder()
        .setRenderUri(renderUri)
        .setMetadata(JSONObject().toString())
        .setAdFilters(filters)
        .setAdCounterKeys(adCounterKeys)
        .build()
    )
  ).build()

Java

// Initialize a custom audience.
CustomAudience audience = new CustomAudience.Builder()
    .setBuyer(buyer)
    .setName(name)
    .setAds(Collections.singletonList(new AdData.Builder()
        .setRenderUri(renderUri)
        .setMetadata(new JSONObject().toString())
        .setAdFilters(filters)
        .setAdCounterKeys(adCounterKeys)
        .build()))
    .build();

When frequency cap filters are implemented into a custom audience, the SSP can then invoke the necessary click, view, or impression events.

Kotlin

val callerAdTech: AdTechIdentifier = mAdSelectionConfig.getSeller()

val request: UpdateAdCounterHistogramRequest = Builder(
  adSelectionId,
  FrequencyCapFilters.AD_EVENT_TYPE_CLICK,  //CLICK, VIEW, or IMPRESSION
  callerAdTech
).build()

Java

AdTechIdentifier callerAdTech = mAdSelectionConfig.getSeller();

UpdateAdCounterHistogramRequest request =
  new UpdateAdCounterHistogramRequest.Builder(
      adSelectionId,
      FrequencyCapFilters.AD_EVENT_TYPE_CLICK, //CLICK, VIEW, or IMPRESSION
      callerAdTech
).build();

Ads that have hit their pre-set frequency cap filter limits are filtered out of the auction. Filtering happens before the bidding logic is executed for on-device auctions, and as the payload is generating for Bidding & Auction services auctions.This toolkit gives ad techs the flexibility to use the interactions between users and the ads within their custom audiences to focus ad targeting while minimizing ad overexposure.

Contextual ad filtering without network calls

If there is no remarketing demand on the device, you can run ad selection for contextual ads without network calls. With prebuilt URIs and a list of contextual ads with bids, the platform can skip retrieving bidding logic, bidding signals, and scoring signals. The platform uses a prebuilt URI to select the contextual ad with the highest bid.

To improve latency, ad techs can run an ad selection flow that includes only contextual ads with ad filtering functionality without network calls. This is achieved by using prebuilt URIs for scoring signals. Refer to the Supported prebuilt URI use cases and names section for a list of scoreAds implementations.

To run ad selection without network calls:

  1. Set up ad filtering
  2. Create your contextual ads
  3. Create a AdSelectionConfig object with the following:

    1. An empty list of buyers
    2. A prebuilt URI to select the highest bid
    3. Contextual ads
    4. An empty URI for the scoring signals. The empty URI is allowed to indicate that you don't want to use the fetching of trusted signals for scoring:
    Uri prebuiltURIScoringUri = Uri.parse("ad-selection-prebuilt://ad-selection/highest-bid-wins/?reportingUrl=your.registered.uri/reporting");
    // Initialize AdSelectionConfig
    AdSelectionConfig adSelectionConfig =
      new AdSelectionConfig.Builder()
        .setSeller(seller)
        .setDecisionLogicUri(prebuiltURIScoringUri)
        .setCustomAudienceBuyers(Collections.emptyList())
        .setAdSelectionSignals(adSelectionSignals)
        .setSellerSignals(sellerSignals)
        .setPerBuyerSignals(perBuyerSignals)
        .setBuyerContextualAds(buyerContextualAds)
        .setTrustedScoringSignalsUri(Uri.EMPTY)
        .build();
    
  4. Run ad selection:

    adSelectionManager.selectAds(
        adSelectionConfig,
        executor,
        outcomeReceiver);
    

Run your own reporting JavaScript while using prebuilt URIs

Today, the Privacy Sandbox platform only has a basic reporting JavaScript implementation available for prebuilt URIs. If you want to run your own reporting JavaScript while still using prebuilt URIs for a low latency ad selection, you can override the DecisionLogicUri between ad selection and reporting runs.

  1. Run steps to run ad selection for contextual ads using prebuilt URIs
  2. Create a copy of your AdSelectionConfig before running reporting

    adSelectionConfigWithYourReportingJS = adSelectionConfig.cloneToBuilder()
      // Replace <urlToFetchYourReportingJS> with your own URL:
      .setDecisionLogicUri(Uri.parse(<urlToFetchYourReportingJS>))
      .build();
    
  3. Run impression reporting

    // adSelectionId is from the result of the previous selectAds run
    ReportImpressionRequest request = new ReportImpressionRequest(
      adSelectionId,
      adSelectionConfigWithYourReportingJS);
    adSelectionManager.reportImpression(
      request,
      executor,
      outcomeReceiver);
    

Run waterfall mediation

Waterfall mediation requires multiple third party SDKs (3P networks) to be orchestrated by a first party SDK mediation network. Waterfall mediation is done the same way regardless if the auction took place on device or ran on Bidding & Auction services (B&A).

3P networks

3P networks need to provide an adapter that allows the mediation network to invoke the necessary methods for running an auction:

  • Run ad selection
  • Report impressions

Here's an example of a mediation network adapter:

Kotlin

class NetworkAdaptor {
    private val adSelectionManager : AdSelectionManager

    init {
        adSelectionManager = context.getSystemService(AdSelectionManager::class.java)
    }

    fun selectAds() {...}

    fun reportImpressions() {...}
}

Java

class NetworkAdaptor {
    AdSelectionManager adSelectionManager;

    public NetworkAdaptor() {
        AdSelectionManager adSelectionManager =
            context.getSystemService(AdSelectionManager.class);
    }

    public void selectAds() {...}

    public void reportImpressions() {...}
}

Each SDK has their own ad selection service managers and clients, and their own selectAds and reportImpressions implementation. SDK providers can refer to the sections on how to run ad selection for on-device auctions or the B&A explainer for B&A auctions. Follow how to report ad impressions (following the single SSP impression reporting for reporting.

Mediation network

Similar to 3P networks, mediation networks need selectAds and reportImpression implementations. Refer to the sections on how to run ad selection and how to report ad impressions for more information.

Mediation networks are responsible for running the mediation chain and placing themselves in the mediation chain. The next section covers how to set up and execute this process.

Retrieve mediation chain and bid floors

The mediation network is responsible for retrieving the first party (1P) contextual ads, mediation chain, and third party networks' bid floors (3P). This can happen in a request to retrieve contextual ads executed by the mediation network. The mediation chain determines how to iterate through the 3P Networks, and the bid floors can be passed to the auction process as adSelectionSignals.

Network placement in the mediation chain

A mediation SDK can place itself in the mediation chain based on their live eCPM of 1P ad bids. In the Protected Audience API, ad bids are opaque. A mediation SDK should use AdSelectionFromOutcomesConfig to be able to compare a given 1P ad's bid to the bid floor of the next 3P network in the chain. If the 1P bid is higher than the bid floor then that means the mediation SDK is placed in front of that 3P network.

Run ad selection

To retrieve a 1P ad candidate, the mediation network can execute an on-device auction following the steps in the run ad selection section. This generates a 1P ad candidate, a bid, and an AdSelectionId that is used in the mediation process.

Create an AdSelectionFromOutcomesConfig

An AdSelectionFromOutcomesConfig allows the mediation network to pass a list of AdSelectionIds (results from previous auctions), ad selection signals, and a URI to fetch JavaScript that selects an ad from multiple candidates. The list of AdSelectionIds along with their bids and the signals are passed to the JavaScript which can return one of the AdSelectionIds if it beats the bid floor, or none if the mediation chain should continued.

Mediation Networks create an AdSelectionFromOutcomesConfig using the 1P AdSelectionId from the previous section, and the bid floor for the 3P Network being considered. A new AdSelectionFromOutcomesConfig should be created for each step in the mediation chain.

Kotlin

fun  runSelectOutcome(
    adSelectionClient : AdSelectionClient,
    outcome1p : AdSelectionOutcome,
    network3p : NetworkAdapter) : ListenableFuture<AdSelectionOutcome?> {
    val config = AdSelectionFromOutcomesConfig.Builder()
        .setSeller(seller)
        .setAdSelectionIds(listOf(outcome1p))
        .setSelectionSignals({"bid_floor": bid_floor})
        .setSelectionLogicUri(selectionLogicUri)
        .build()
    return adSelectionClient.selectAds(config)
}

Java

public ListenableFuture<AdSelectionOutcome> runSelectOutcome(AdSelectionOutcome outcome1p,
                                              NetworkAdapter network3p) {
    AdSelectionFromOutcomesConfig config = new AdSelectionFromOutcomesConfig.Builder()
            .setSeller(seller)
            .setAdSelectionIds(Collection.singletonList(outcome1p))
            .setSelectionSignals({"bid_floor": bid_floor})
            .setSelectionLogicUri(selectionLogicUri)
            .build();

    return adSelectionClient.selectAds(config){}
}

The selectAds() method override for waterfall mediation requires an AdSelectionFromOutcomesConfig input, where you must specify the following required parameters:

  • Seller: Identifier for the seller ad network initiating the ad selection.
  • AdSelectionIds: A singleton list of a previous selectAds() run for a 1P ad.
  • Ad selection signals: A JSON object, serialized as a string, containing signals to be used by buyer bidding logic. In this case, include the bid floor retrieved for the given 3P network.
  • Selection Logic URI: An HTTPS URL queried during ad selection to fetch the mediation network's JavaScript for selecting a winning ad. See the required function signatures in this JavaScript. The JavaScript should return the 3P ad if the bid is higher than the bid floor, or otherwise return null. This allows the mediation SDK to truncate the mediation chain when a winner is found.

With the AdSelectionOutcomesConfig created, call the selectAds() method of the 3P network that is first in the chain.

Kotlin

val adSelectionManager = context.getSystemService(AdSelectionManager::class.java)

// Initialize AdSelectionFromOutcomesConfig
AdSelectionFromOutcomesConfig adSelectionFromOutcomesConfig =
  AdSelectionFromOutcomesConfig.Builder()
    .setSeller(seller)
    .setAdSelectionIds(listof(outcome1p))
    .setSelectionSignals({"bid_floor": bid_floor})
    .setSelectionLogicUri(selectionLogicUri)
    .setAdSelectionIds(outcomeIds)
    .build()

// Run ad selection with AdSelectionConfig
adSelectionManager.selectAds(
    adSelectionFromOutcomesConfig,
    executor,
    outcomeReceiver)

Java

AdSelectionManager adSelectionManager =
    context.getSystemService(AdSelectionManager.class);

// Initialize AdSelectionFromOutcomesConfig
AdSelectionFromOutcomesConfig adSelectionFromOutcomesConfig =
        new AdSelectionFromOutcomesConfig.Builder()
            .setSeller(seller)
            .setAdSelectionIds(Collection.singletonList(outcome1p))
            .setSelectionSignals({"bid_floor": bid_floor})
            .setSelectionLogicUri(selectionLogicUri)
            .setAdSelectionIds(outcomeIds)
            .build();

// Run ad selection with AdSelectionConfig
adSelectionManager.selectAds(
    adSelectionFromOutcomesConfig,
    executor,
    outcomeReceiver);

Orchestrate waterfall mediation

The following is the order of operations for running through the mediation process.

  1. Run 1P ad selection.
  2. Iterate over the mediation chain. For each 3P network, do the following:
    1. Build AdSelectionFromOutcomeConfig including the 1P outcomeId and the 3P SDK's bid floor.
    2. Call selectAds() with the config from the previous step.
    3. If the result is not empty, return the ad.
    4. Call the current SDK network adapter's selectAds() method. If the result is not empty, return the ad.
  3. If no winner is found from the chain, return the 1P ad.

Kotlin

fun runWaterfallMediation(mediationChain : List<NetworkAdapter>)
  : Pair<AdSelectionOutcome, NetworkAdapter> {
    val outcome1p = runAdSelection()

    var outcome : AdSelectionOutcome
    for(network3p in mediationChain) {
      outcome = runSelectOutcome(outcome1p, network3p)
      if (outcome1p.hasOutcome() && outcome.hasOutcome()) {
          return Pair(outcome, this)
      }

      outcome = network3p.runAdSelection()
      if(outcome.hasOutcome()) {
          return Pair(outcome, network3p)
      }
    }
  return Pair(outcome1p, this)
}

Java

class MediationNetwork {
    AdSelectionManager adSelectionManager;

    public MediationNetwork() {
        AdSelectionManager adSelectionManager =
            context.getSystemService(AdSelectionManager.class);
    }

    public void runAdSelection() {...}

    public void reportImpressions() {...}

    public Pair<AdSelectionOutcome, NetworkAdapter> runWaterfallMediation(
            List<NetworkAdapter> mediationChain) {
        AdSelectionOutcome outcome1p = runAdSelection();

        AdSelectionOutcome outcome;
        for(NetworkAdapter network3p: mediationChain) {
            if (outcome1p.hasOutcome() &&
              (outcome = runSelectOutcome(outcome1p, network3p)).hasOutcome()) {
                return new Pair<>(outcome, this);
            }

            if((outcome = network3p.runAdSelection()).hasOutcome()) {
                return new Pair<>(outcome, network3p);
            }
        }
        return new Pair<>(outcome1p, this);
    }

    /* Runs comparison by creating an AdSelectionFromOutcomesConfig */
    public AdSelectionOutcome runSelectOutcome(AdSelectionOutcome outcome1p,
                                              NetworkAdapter network3p) { ... }
}

Report ad impressions

There are two flows for reporting an ad impression depending on how the auction is run. If you are a single SSP running an auction, follow this section. If you are going to implement waterfall mediation, follow the steps found in the waterfall mediation impression reporting section.

Single SSP impression reporting

After a winning ad has been chosen from the ad selection workflow, you can report the impression back to participating buy-side and sell-side platforms with the AdSelectionManager.reportImpression() method. To report an ad impression:

  1. Initialize an AdSelectionManager object.
  2. Build a ReportImpressionRequest object with the ad selection ID.
  3. Call the asynchronous reportImpression() method with the ReportImpressionRequest object and relevant Executor and OutcomeReceiver objects.

Java

AdSelectionManager adSelectionManager =
    context.getSystemService(AdSelectionManager.class);

// Initialize a ReportImpressionRequest
ReportImpressionRequest reportImpressionRequest =
        new ReportImpressionRequest.Builder()
                .setAdSelectionId(adSelectionId)
                .setAdSelectionConfig(adSelectionConfig)
                .build();

// Request to report the impression with the ReportImpressionRequest
adSelectionManager.reportImpression(
    reportImpressionRequest,
    executor,
    outcomeReceiver);

Kotlin

val adSelectionManager = context.getSystemService(AdSelectionManager::class.java)

// Initialize a ReportImpressionRequest
val adSelectionConfig: ReportImpressionRequest =
    ReportImpressionRequest.Builder()
        .setAdSelectionId(adSelectionId)
        .setAdSelectionConfig(adSelectionConfig)
        .build()

// Request to report the impression with the ReportImpressionRequest
adSelectionManager.reportImpression(
    reportImpressionRequest,
    executor,
    outcomeReceiver)

Initialize the ReportImpressionRequest with the following required parameters:

  • Ad selection ID: An ID unique only to a device user that identifies a successful ad selection.
  • Ad selection config: The same configuration used in the selectAds() call identified by the provided ad selection ID.

The asynchronous reportImpression() method uses the OutcomeReceiver object to signal the result of the API call.

  • The onResult() callback indicates if impression reporting URLs have been created and the request has been scheduled.
  • The onError() callback indicates the following possible conditions:
    • If the call is initialized with an invalid input argument, the AdServicesException indicates an IllegalArgumentException as the cause.
    • All other errors receive an AdServicesException with an IllegalStateException as the cause.

Waterfall mediation impression reporting

A mediation SDK needs to keep track of the winning SDK to trigger their reporting flows. The SDKs participating in a mediation chain should provide a method for the mediator to invoke to trigger their own reporting flow. An SDK participating in a mediated auction can follow the steps above to implement their own reporting.

SSPs can use this 3P SDK code example as a prototype for how to join in mediation flows:

Pair<AdSelectionOutcome, NetworkAdapter> winnerOutcomeAndNetwork =
         mediationSdk.orchestrateMediation(mediationChain);

if (winner.first.hasOutcome()) {
      winner.second.reportImpressions(winner.first.getAdSelectionId());

Impression reporting endpoints

The report impression API issues HTTPS GET requests to endpoints provided by the sell-side platform and the winning buy-side platform:

Buy-side platform endpoint:

  • The API uses the Bidding logic URL specified in the custom audience to fetch the buyer-provided JavaScript that includes logic to return an impression reporting URL.
  • Invoke the reportWin() JavaScript function, which is expected to return the buyer's impression reporting URL.

Sell-side platform endpoint:

  • Use the Decision logic URL specified in the AdSelectionConfig object to fetch the seller's decision logic JavaScript.
  • Invoke the reportResult() JavaScript function, which is expected to return the seller's impression reporting URL.

Bidding & Auction services reporting

An auction executed on Bidding & Auction services will have all the necessary reporting information, including generated URLs for ad interaction reporting, included in the encrypted response from the server-side auction. When the response is decrypted, the appropriate URLs are registered with the platform, so ad and impression reporting follows the same steps listed above.

Best effort Impression reporting

The reportImpression() method is designed to offer a best-effort completion of reporting.

Report Ad Interactions

Protected Audience provides support to report on more granular interactions for a rendered ad. This can include interactions such as view time, clicks, hovers, or any other useful metric that can be collected. The process to receive these reports requires two steps. First, buyers and sellers must register to receive these reports in their reporting JavaScript. Then, the client will need to report these events.

Registering to receive interaction events

Registering for interaction events happens in the buyer's reportWin() and seller's reportResult() JavaScript functions using a JavaScript function provided by the platform: registerAdBeacon. To register to receive an event report, simply call the platform JavaScript Function from your reporting JavaScript. The following snippet is using a buyer's reportWin(), but the same approach applies to reportResult().

reportWin(
  adSelectionSignals,
  perBuyerSignals,
  signalsForBuyer,
  contextualSignals,
  customAudienceSignals) {
    ...
    // Calculate reportingUri, clickUri, viewUri, and hoverUri

    registerAdBeacon({"click": clickUri, "view": viewUri, "hover": hoverUri});

    return reportingUri;
}

Reporting interaction events

After reporting an impression, clients can report the interactions back to previously registered winning buy-side and sell-side platforms with the AdSelectionManager.reportInteraction() method. To report an ad event:

  1. Initialize an AdSelectionManager object.
  2. Build a ReportInteractionRequest object with the ad selection ID, interaction key, interaction data, and reporting destination.
  3. Call the asynchronous reportInteraction() method with the request object and relevant Executor and OutcomeReceiver objects.
AdSelectionManager adSelectionManager =
    context.getSystemService(AdSelectionManager.class);

// Initialize a ReportInteractionRequest
ReportInteractionRequest request =
  new ReportInteractionRequest.Builder()
    .setAdSelectionId(adSelectionId)
    .setInteractionKey("view")
    .setInteractionData("{ viewTimeInSeconds : 1 }") // Can be any string
    .setReportingDestinations(
      FLAG_REPORTING_DESTINATION_BUYER | FLAG_REPORTING_DESTINATION_SELLER
    )
    .build();

// Request to report the impression with the ReportImpressionRequest
adSelectionManager.reportInteraction(
  reportImpressionRequest,
  executor,
  outcomeReceiver);

Initialize the ReportInteractionRequest with the following required parameters:

  • Ad selection ID: An ad selection ID retrieved from a previously returned AdSelectionOutcome.
  • Interaction Key: A string key defined by the client describing the action being reported. This must match the key that was registered by the seller or buyer in the reporting JavaScript functions.
  • Interaction Data: A string containing data to be included with the event report, to be POSTed back to the reporting servers.
  • Reporting Destinations: A bit mask specifying if the events should be reported to the buyer, seller, or both. These flags are provided by the platform and the final destination mask can be created using bitwise operations. To report to one destination, you can use the flag provided by the platform directly. To report to multiple destinations, you can use the bitwise OR (|) to combine flag values.

The asynchronous reportInteraction() method uses the OutcomeReceiver object to signal the result of the API call.

  • The onResult() callback indicates the report interaction call is valid.
  • The onError() callback indicates the following possible conditions:
    • If the call is made when the app is running in the background, an IllegalStateException with a description of the failure is returned.
    • If the client is throttled from calling reportInteraction(), a LimitExceededException is returned.
    • If the package is not enrolled to call the Privacy Preserving APIs, a SecurityException() is returned.
    • If the app reporting interactions is different from the app that called selectAds(), an IllegalStateException is returned.
  • If the user has not consented to enable the Privacy Sandbox APIs, the call will fail silently.

Interaction reporting endpoints

The report interaction API issues HTTPS POST requests to endpoints provided by the sell-side platform and the winning buy-side platform. Protected Audience will match the interaction keys with the URIs declared in reporting JavaScript and issue a POST request to each endpoint for each interaction being reported. The content type of the request is plain text with the body being the Interaction Data.

Best effort Interaction reporting

The reportInteraction() is designed to offer a best-effort completion of reporting through HTTP POST.

Daily background update

When creating a custom audience, your app or SDK can initialize custom audience metadata. Additionally, the platform can update the following pieces of custom audience metadata with a daily background update process.

  • User bidding signals
  • Trusted bidding data
  • AdData list

This process queries against the Daily update URL defined in the custom audience and the URL may return a JSON response.

  • The JSON response may contain any of the supported metadata fields that needs to be updated.
  • Each JSON field is validated independently. The client ignores any malformed fields which results in no updates to that particular field in the response.
  • An empty HTTP response or an empty JSON object "{}" results in no metadata updates.
  • The response message size must be limited to 10 KB.
  • All URIs are required to use HTTPS.
  • trusted_bidding_uri must share the same ETLD+1 as the buyer.

Example: JSON response for background daily update

{
    "user_bidding_signals" : { ... },  // Valid JSON object
    "trusted_bidding_data" : {
        "trusted_bidding_uri" : 'example-dsp1-key-value-service.com',
        "trusted_bidding_keys" : [ 'campaign123', 'campaign456', ... ]
    },
    'ads' : [
        {
            "render_uri" : 'www.example-dsp1.com/.../campaign123.html',
            'metadata' : { ... }  // Valid JSON object
        },
        {
            "render_uri" : 'www.example-dsp1.com/.../campaign456.html',
            'metadata' : { ... }  // Valid JSON object
        },
        ...
    ]
}

JavaScript for ad selection

The ad selection workflow orchestrates the execution of buyer-provided and seller-provided JavaScript.

Buyer-provided JavaScript is fetched from the Bidding logic URL specified in the custom audience. The returned JavaScript should include the following functions:

Seller-provided JavaScript is fetched from the decision logic URL specified in the AdSelectionConfig parameter for the ad selection API. The returned JavaScript should include the following functions:

generateBid()

function generateBid(
  ad,
  auction_signals,
  per_buyer_signals,
  trusted_bidding_signals,
  contextual_signals,
  user_signals,
  custom_audience_bidding_signals) {
  return {'status': 0, 'ad': ad, 'bid': ad.metadata.result };
}

Input parameters:

  • ad: a JSON object with the format var ad = { 'render_url': url, 'metadata': json_metadata };
  • auction_signals, per_buyer_signals: JSON objects specified in the auction configuration object
  • custom_audience_bidding_signals: JSON object generated by the platform. The format for this JSON object is:

    var custom_audience_signals = {
      "owner":"ca_owner",
      "buyer":"ca_buyer",
      "name":"ca_name",
      "activation_time":"ca_activation_time_epoch_ms",
      "expiration_time":"ca_expiration_time_epoch_ms",
      "user_bidding_signals":"ca_user_bidding_signals"
    }
    

    where:

    • owner, buyer, and name are string taken from the properties with the same name of the Custom Audience participating to the ad selection
    • activation_time and expiration_time are the time of activation and expiration of the custom audience, expressed as seconds since the Unix epoch
    • ca_user_bidding_signals is a JSON string specified in the userBiddingSignals field of the CustomAudience at creation time
    • trusted_bidding_signals, contextual_signals, and user_signals are JSON objects. They are passed as empty objects and will be filled up in future releases. Their format is not enforced by the platform and is managed by the ad tech.

Result:

  • ad: is the ad the bid refers to. The script is allowed to return a copy of the ad it received with different metadata. The render_url property of the ad is expected to be unaltered.
  • bid: a float value representing the bid value for this ad
  • status: an integer value that can be:
    • 0: for a successful execution
    • 1: (or any non-zero value) in case any of the input signals is invalid. In case a non-zero value is returned by generate-bid the bidding process is invalidated for all the CA ads

scoreAd()

function scoreAd(
  ad,
  bid,
  ad_selection_config,
  seller_signals,
  trusted_scoring_signals,
  contextual_signal,
  user_signal,
  custom_audience_signal) {
    return {'status': 0, 'score': score };
}

Input parameters:

  • ad: see the generateBid documentation
  • bid: the bid value for the ad
  • ad_selection_config: a JSON object representing the AdSelectionConfig parameter of the selectAds API. The format is:

    var ad_selection_config = {
      'seller': 'seller',
      'decision_logic_url': 'url_of_decision_logic',
      'custom_audience_buyers': ['buyer1', 'buyer2'],
      'auction_signals': auction_signals,
      'per_buyer_signals': per_buyer_signals,
      'contextual_ads': [ad1, ad2]
    }
    
  • seller_signals: JSON objects read from the sellerSignals AdSelectionConfig API parameter

  • trusted_scoring_signal: read from the adSelectionSignals field in the AdSelectionConfig API parameter

  • contextual_signals, user_signals: JSON objects. They are currently passed as empty objects and will be filled up in future releases. Their format is not enforced by the platform and is managed by the ad tech.

  • per_buyer_signals: JSON object read from the perBuyerSignal map in the AdSelectionConfig API parameter using as key the current Custom Audience buyer. Empty if the map doesn't contain any entry for the given buyer.

Output:

  • score: a float value representing the score value for this ad
  • status: an integer value that can be:
    • 0: for a successful execution
    • 1: in case the customAudienceSignals are invalid
    • 2: in case the AdSelectionConfig is invalid
    • 3: in case any of the other signals is invalid
    • Any non-zero value causes the failure of the process, the value determines the type of exception thrown

selectOutcome()

function selectOutcome(
  outcomes,
  selection_signals) {
    return {'status': 0, 'result': null};
}

Input parameters:

  • outcomes: a JSON object {"id": id_string, "bid": bid_double}
  • selection_signals: JSON objects specified in the auction configuration object

Output:

  • status: 0 for success, non-zero for failure
  • result: one of the outcomes passed in or null

reportResult()

function reportResult(ad_selection_config, render_url, bid, contextual_signals) {
   return {
      'status': status,
      'results': {'signals_for_buyer': signals_for_buyer, 'reporting_url': reporting_url }
   };
}

Input parameters:

  • ad_selection_config: see the documentation of scoreAds
  • render_url: the render URL of the winning ad
  • bid: the bid offered for the winning ad
  • contextual_signals: see the documentation of generateBid

Output:

  • status: 0 for success and non-zero for failure
  • results: a JSON objects containing:
    • signals_for_buyer: a JSON object that is passed to the reportWin function
    • reporting_url: a URL that is used by the platform to notify the impression to the buyer

reportWin()

function reportWin(
   ad_selection_signals,
   per_buyer_signals,
   signals_for_buyer,
   contextual_signals,
   custom_audience_signals) {
   return {'status': 0, 'results': {'reporting_url': reporting_url } };
}

Input parameters:

  • ad_selection_signals, per_buyer_signals: see the documentation for scoreAd
  • signals_for_buyer: a JSON object returned by reportResult
  • contextual_signals, custom_audience_signals: see the documentation for generateBid

Output:

  • status: 0 for success and non-zero for failure
  • results: a JSON object containing:
    • reporting_url: a URL that is used by the platform to notify the impression to the seller

registerAdBeacon()

function registerAdBeacon(
  beacons
)

Input Parameters:

  • beacons: An object containing key-value pairs of interaction keys and reporting URIs. The format is:

    let beacons = {
      'interaction_key': 'reporting_uri',
      'interaction_key': 'reporting_uri',
      ...
    }
    
    • interaction_key: A string representing the event. This is used by the platform later when reporting event interactions to look up the reporting_uri that should be notified. This key needs to match between what the buyer or seller is registering, and what the seller is reporting.
    • reporting_uri: A URI to receive event reports. This should be specific to the event type being reported. It must accept a POST request to handle any data reported along with the event.

    For example:

      let beacons = {
        'click': 'https://reporting.example.com/click_event',
        'view': 'https://reporting.example.com/view_event'
      }
    

Ad Selection prebuilt URIs

Prebuilt URIs give ad techs the ability to appoint JavaScript functions for ad selection decision logic in the AdSelectionConfig and AdSelectionFromOutcomesConfig classes. Prebuilt URIs don't require network calls to download the corresponding JavaScript. Ad techs can use prebuilt URIs without having to set up an enrolled domain to host the JavaScript.

A prebuilt URI is constructed using the following format:

ad-selection-prebuilt:<use-case>/<name>?<required-script-generation-parameters>

The Privacy Sandbox platform provides JavaScript using the information from this URI in the runtime.

An IllegalArgumentException is thrown if:

  • any of the required parameters are not present in the URI
  • there are unrecognized parameters in the URI

Supported prebuilt URI use cases and names

Use case 1: ad-selection

Prebuilt URIs under the ad-selection use case are supported in the selectAds(AdSelectionConfig) flow.

Prebuilt URI name: highest-bid-wins

This prebuilt URI provides a JavaScript that picks the ad with the highest bid after bidding. It also provides a basic reporting function to report the winner's render_uri and bid.

Required parameters

reportingUrl: The base reporting URL that is parameterized with the render_uri and the bid of the winning ad:

<reportingUrl>?render_uri=<renderUriOfWinnigAd>&bid=<bidOfWinningAd>

Usage

If your base reporting URL is https://www.ssp.com/reporting then the prebuilt URI would be:

`ad-selection-prebuilt://ad-selection/highest-bid-wins/?reportingUrl=https://www.ssp.com/reporting`

Use case 2: ad-selection-from-outcomes

Prebuilt URIs under the ad-selection-from-outcomes use case support the selectAds(AdSelectionFromOutcomesConfig) workflow.

Prebuilt URI name: waterfall-mediation-truncation

The waterfall-mediation-truncation prebuilt URI provides JavaScript that implements waterfall mediation truncation logic where the JavaScript returns a first-party ad if the bid is higher then or equal to the bid floor, and otherwise returns null.

Required parameters

bidFloor: The key of the bid floor value passed in the getSelectionSignals() that is compared against the mediation SDK's ad.

Usage

If your ad selection signals look like {"bid_floor": 10} then the resulting prebuilt URI would be:

`ad-selection-prebuilt://ad-selection-from-outcomes/waterfall-mediation-truncation/?bidFloor=bid_floor`

Testing

To help you get started with the Protected Audience API, we've created sample apps in Kotlin and Java, which can be found on GitHub.

Prerequisites

The Protected Audience API requires some JavaScript during ad selection and impression reporting. There are two methods of providing this JavaScript in a testing environment:

  • Run a server with the required HTTPS endpoints that returns the JavaScript
  • Override remote fetching by providing the necessary code from a local source

Either approach requires setting up an HTTPS endpoint to handle impression reporting.

HTTPS endpoints

To test ad selection and impression reporting, you need to set up 7 HTTPS endpoints that your test device or emulator can access:

  1. Buyer endpoint that serves the bidding logic JavaScript.
  2. An endpoint that serves the bidding signals.
  3. Seller endpoint that serves the decision logic JavaScript.
  4. An endpoint that serves scoring signals.
  5. Winning buyer impression reporting endpoint.
  6. Seller impression reporting endpoint.
  7. An endpoint to serve the daily updates for a custom audience.

For convenience, the GitHub repo provides basic JavaScript code for testing purposes. It also includes OpenAPI service definitions which can be deployed to a supported mock or microservices platform. For more details, see the project README.

Override remote fetching of JavaScript

This feature is intended to be used for end-to-end testing. To override remote fetching, your app must run in debug mode with developer options enabled.

To enable debug mode for your application, add the following line to the application attribute in your AndroidManifest.xml:

<application
  android:debuggable="true">

For an example of how to use these overrides, please see the the Protected Audience API sample app on GitHub.

You need to add your own custom JavaScript to handle ad selection routines such as bidding, scoring decisions, and reporting. You can find basic JavaScript code examples that handle all required requests in the GitHub repo. The Protected Audience API sample application demonstrates how to read code from that file and prepare it for use as an override.

It is possible to override sell-side and buy-side JavaScript fetching independently, though you need an HTTPS endpoint to serve any JavaScript you aren't providing overrides for. Please see the README for information about how to set up a server that handles these cases.

It is only possible to override JavaScript fetching for custom audiences that are owned by your package.

Override sell-side JavaScript

To set up an override of sell-side JavaScript, do the following as demonstrated in the following code example:

  1. Initialize an AdSelectionManager object.
  2. Get a reference to TestAdSelectionManager from the AdSelectionManager object.
  3. Build an AdSelectionConfig object.
  4. Build an AddAdSelectionOverrideRequest with the AdSelectionConfig object and a String representing the JavaScript you intend to use as an override.
  5. Call the asynchronous overrideAdSelectionConfigRemoteInfo() method with the AddAdSelectionOverrideRequest object and relevant Executor and OutcomeReceiver objects.

Kotlin

val testAdSelectionManager: TestAdSelectionManager =
  context.getSystemService(AdSelectionManager::class.java).getTestAdSelectionManager()

// Initialize AdSelectionConfig =
val adSelectionConfig = new AdSelectionConfig.Builder()
    .setSeller(seller)
    .setDecisionLogicUrl(decisionLogicUrl)
    .setCustomAudienceBuyers(customAudienceBuyers)
    .setAdSelectionSignals(adSelectionSignals)
    .setSellerSignals(sellerSignals)
    .setPerBuyerSignals(perBuyerSignals)
    .build()

// Initialize AddAddSelectionOverrideRequest
val request = AddAdSelectionOverrideRequest.Builder()
    .setAdSelectionConfig(adSelectionConfig)
    .setDecisionLogicJs(decisionLogicJS)
    .build()

// Run the call to override the JavaScript for the given AdSelectionConfig
// Note that this only takes effect in apps marked as debuggable
testAdSelectionManager.overrideAdSelectionConfigRemoteInfo(
    request,
    executor,
    outComeReceiver)

Java

TestAdSelectionManager testAdSelectionManager =
  context.getSystemService(AdSelectionManager.class).getTestAdSelectionManager();

// Initialize AdSelectionConfig =
AdSelectionConfig adSelectionConfig = new AdSelectionConfig.Builder()
    .setSeller(seller)
    .setDecisionLogicUrl(decisionLogicUrl)
    .setCustomAudienceBuyers(customAudienceBuyers)
    .setAdSelectionSignals(adSelectionSignals)
    .setSellerSignals(sellerSignals)
    .setPerBuyerSignals(perBuyerSignals)
    .build();

// Initialize AddAddSelectionOverrideRequest
AddAdSelectionOverrideRequest request = AddAdSelectionOverrideRequest.Builder()
    .setAdSelectionConfig(adSelectionConfig)
    .setDecisionLogicJs(decisionLogicJS)
    .build();

// Run the call to override the JavaScript for the given AdSelectionConfig
// Note that this only takes effect in apps marked as debuggable
testAdSelectionManager.overrideAdSelectionConfigRemoteInfo(
    request,
    executor,
    outComeReceiver);

See the Run ad selection section for more information about what each of the fields in the AdSelectionConfig represent. The key difference is that the decisionLogicUrl can be set to a placeholder value as it will be ignored.

In order to override the JavaScript used during ad selection, the decisionLogicJs must contain the proper seller-side function signatures. For an example of how to read a JavaScript file as a string, please see the Protected Audience API sample app on GitHub.

The asynchronous overrideAdSelectionConfigRemoteInfo() method uses the OutcomeReceiver object to signal the result of the API call.

The onResult() callback signifies the override was applied successfully. Future calls to selectAds() will use whatever decision and reporting logic you have passed in as the override.

The onError() callback signifies two possible conditions:

  • If the override is attempted with invalid arguments, the AdServiceException indicates an IllegalArgumentException as the cause.
  • If the override is attempted with an app not running in debug mode with developer options enabled, the AdServiceException indicates IllegalStateException as the cause.

Reset sell-side overrides

This section assumes that you have overridden the sell-side JavaScript and that you have a reference to the TestAdSelectionManager and AdSelectionConfig used in the previous section.

In order to reset the overrides for all AdSelectionConfigs:

  1. Call the asynchronous resetAllAdSelectionConfigRemoteOverrides() method with the relevant OutcomeReceiver object.

Kotlin

// Resets overrides for all AdSelectionConfigs
testAadSelectionManager.resetAllAdSelectionConfigRemoteOverrides(
  outComeReceiver)

Java

// Resets overrides for all AdSelectionConfigs
testAdSelectionManager.resetAllAdSelectionConfigRemoteOverrides(
    outComeReceiver);

After you reset sell-side overrides, calls to selectAds() use whatever decisionLogicUrl is stored in the AdSelectionConfig to attempt to fetch the required JavaScript.

If the call to resetAllAdSelectionConfigRemoteOverrides() fails, the OutComeReceiver.onError() callback provides an AdServiceException. If the removal of overrides is attempted with an app not running in debug mode with developer options enabled, AdServiceException indicates IllegalStateException as the cause.

Override buy-side JavaScript

  1. Follow the steps to join a custom audience
  2. Build an AddCustomAudienceOverrideRequest with the buyer and name of the custom audience you wish to override, in addition to the bidding logic and data you wish to use as an override
  3. Call the asynchronous overrideCustomAudienceRemoteInfo() method with the AddCustomAudienceOverrideRequest object and relevant Executor and OutcomeReceiver objects

Kotlin

val testCustomAudienceManager: TestCustomAudienceManager =
  context.getSystemService(CustomAudienceManager::class.java).getTestCustomAudienceManager()

// Join custom audience

// Build the AddCustomAudienceOverrideRequest
val request = AddCustomAudienceOverrideRequest.Builder()
    .setBuyer(buyer)
    .setName(name)
    .setBiddingLogicJs(biddingLogicJS)
    .setTrustedBiddingSignals(trustedBiddingSignals)
    .build()

// Run the call to override JavaScript for the given custom audience
testCustomAudienceManager.overrideCustomAudienceRemoteInfo(
    request,
    executor,
    outComeReceiver)

Java

TestCustomAudienceManager testCustomAudienceManager =
  context.getSystemService(CustomAudienceManager.class).getTestCustomAudienceManager();

// Join custom audience

// Build the AddCustomAudienceOverrideRequest
AddCustomAudienceOverrideRequest request =
    AddCustomAudienceOverrideRequest.Builder()
        .setBuyer(buyer)
        .setName(name)
        .setBiddingLogicJs(biddingLogicJS)
        .setTrustedBiddingSignals(trustedBiddingSignals)
        .build();

// Run the call to override JavaScript for the given custom audience
testCustomAudienceManager.overrideCustomAudienceRemoteInfo(
    request,
    executor,
    outComeReceiver);

The values for buyer and name are the same ones used to create the custom audience. Learn more about these fields.

Additionally, you can specify two additional parameters:

  • biddingLogicJs: JavaScript that holds the buyer's logic that is used during ad selection. See the required function signatures in this JavaScript.
  • trustedBiddingSignals: Bidding signals to be used during ad selection. For testing purposes this can be an empty string.

The asynchronous overrideCustomAudienceRemoteInfo() method uses the OutcomeReceiver object to signal the result of the API call.

The onResult() callback signifies the override was applied successfully. Subsequent calls to selectAds() use whatever bidding and reporting logic you have passed in as the override.

The onError() callback signifies two possible conditions.

  • If the override is attempted with invalid arguments, the AdServiceException indicates an IllegalArgumentException as the cause.
  • If the override is attempted with an app not running in debug mode with developer options enabled, the AdServiceException indicates IllegalStateException as the cause.

Reset buy-side overrides

This section assumes that you have overridden the buy-side JavaScript and that you have a reference to the TestCustomAudienceManager used in the previous section.

To reset overrides for all custom audiences:

  1. Call the asynchronous resetAllCustomAudienceOverrides() method with relevant Executor and OutcomeReceiver objects.

Kotlin

// Resets overrides for all custom audiences
testCustomAudienceManager.resetCustomAudienceRemoteInfoOverride(
    executor,
    outComeReceiver)

Java

// Resets overrides for all custom audiences
testCustomAudienceManager.resetCustomAudienceRemoteInfoOverride(
    executor,
    outComeReceiver)

After you reset buy-side overrides, subsequent calls to selectAds() use whatever biddingLogicUrl and trustedBiddingData is stored in the CustomAudience to attempt to fetch the required JavaScript.

If the call to resetCustomAudienceRemoteInfoOverride() fails, the OutComeReceiver.onError() callback provides an AdServiceException. If the removal of overrides is attempted with an app not running in debug mode with developer options enabled, the AdServiceException indicates IllegalStateException as the cause.

Set Up a Reporting Server

When you use remote fetching overrides, you'll still need to set up a server that your device or emulator can reach to respond to reporting events. A simple endpoint that returns 200 is sufficient for testing. The GitHub repo includes OpenAPI service definitions which can be deployed to a supported mock or microservices platform. For more details, see the project README.

When looking for the OpenAPI definitions, look for the reporting-server.json. This file contains a simple endpoint that returns 200, representing an HTTP response code. This endpoint is used during selectAds() and signals to the Protected Audience API that impression reporting completed successfully.

Functionality to test

  • Exercise joining or leaving and setting up a custom audience based on prior user actions.
  • Exercise the initiation of on-device ad selection through JavaScripts hosted remotely.
  • Observe how an app's association with custom audience settings may affect ad selection outcomes.
  • Exercise impression reporting after ad selection.

Limitations

The following table lists limitations for the Protected Audience API processing. The limits presented could be subject to change based on feedback. For in-progress capabilities, read the release notes.

Component Limit Description Limit Value
Custom audience (CA) Maximum number of ads per CA 100
Maximum number of CAs per application 1000
Maximum number of apps that can create a CA 1000
Maximum delay in the activation time of a CA from its creation time 60 days
Maximum expiration time of a CA from its activation time 60 days
Maximum number of CAs on device 4000
Maximum size of CA name 200 bytes
Maximum size of daily fetch URI 400 bytes
Maximum size of bidding logic URI 400 bytes
Maximum size of trusted bidding data 10 KB
Maximum size of user bidding signals 10 KB
Maximum call rate for leaveCustomAudience per buyer 1 per second
Maximum call rate for joinCustomAudience per buyer 1 per second
CA Background Fetch Connect timeout 5 seconds
HTTP read timeout 30 seconds
Maximum total download size 10 KB
Max duration of a fetch iteration 5 minutes
Maximum number of CAs updated per job 1000
Ad Selection Maximum number of buyers TBD
Maximum number of CAs per buyer TBD
Maximum number of ads in an auction TBD
Initial connection timeout 5 seconds
Connection read timeout 5 seconds
Maximum execution time of overall AdSelection 10 seconds
Maximum execution time of bidding per CA in AdSelection 5 second
Maximum execution time of scoring in AdSelection 5 second
Maximum execution time for per buyer in AdSelection TBD
Maximum size of ad selection/seller/per buyer signals TBD
Maximum size of seller/buyer scripts TBD
Maximum call rate for selectAds 1 QPS
Impression reporting Minimum time before removing ad selection from persistence 24 hrs
Maximum number of storage ad selections TBD
Maximum size of reporting output URL TBD
Maximum time for impression reporting TBD
Maximum number of retries for notification calls TBD
Connection timeout 5 seconds
Maximum overall execution time for reportImpression 2 seconds
Maximum call rate for reportImpressions 1 QPS
Event reporting Maximum number of beacons per buyer per auction 10

Maximum number of beacons per seller per auction

10

Maximum size of event key

40 bytes

Maximum size of event data

64KB

Ads Maximum size of ad list 10 KB shared by all AdData in a single CA for contextual
URLs Maximum length of any URL string taken as input TBD
Javascript Maximum execution time 1 second for bidding and scoring for impression reporting
Maximum memory used 10 MB

Report bugs and issues

Your feedback is a crucial part of the Privacy Sandbox on Android! Let us know of any issues you find or ideas for improving Privacy Sandbox on Android.